Metagenomic analysis of gut microbiome and resistome of diarrheal fecal samples from Kolkata, India, reveals the core and variable microbiota including signatures of microbial dark matter

Diarrhea is a leading cause of mortality accounting for more than 1.6 million deaths worldwide [1]. It causes nearly 5,25,000 deaths among children under 5 years of age and leads to malnutrition, stunted growth and anemia [2‐7]. It is particularly prevalent in the low and middle-income countries owing to poor hygiene and sanitation. India is the second most populous country in the world and is one of the top five countries with the highest burden of diarrhea and high rates of mortality and morbidity [8‐10]. Recently, India has recorded the highest number of deaths among under five age group [11]. The Eastern region recorded the third highest mortality rate among under five age and diarrhea is one of the leading causes of death in this region [11]. In India the most common causes of diarrhea are Rotavirus, Cryptosporidium sp. Shigella sp., Enterotoxigenic Eschericia coli [12, 16]. Antibiotic therapy is administered to diarrheal patients along with ORS (oral rehydration solution) to assuage severity of symptoms. AMR (antimicrobial resistance) has rendered antibiotic therapy in diarrhea partially or completely ineffective. The genetic determinants of AMR reside in the gut and in the environmental microbiota from where they spread and enter into diarrheal pathogens by lateral gene transfer (LGT). Most of the diarrheal pathogens like E.coli, Klebsiella pneumoniae, Campylobacter sp., Shigella sp. have emerged as multidrug-resistant (MDR) and extensively drug-resistant (XDR) and fail to respond to empirical drugs like aminoglycosides and cephalosporin [13]. AMR is a global challenge which needs to be urgently addressed using a multi-disciplinary approach. Surveillance of AMR in diarrheal patients based on next-generation sequencing (NGS) is a novel way of addressing the AMR threat [13]. The structural and functional components of the microbiota can be studied and mapped completely with the aid of culture-free techniques which have been possible due to the advent of NGS. Big data derived from sequencing metagenomes will help to understand the importance of the structural and functional components of the microbiota in the development and dissemination of AMR [13] by detection, analysis of distribution and abundance of AMR determinants and their source organisms. A large number of studies have been undertaken over the last decades to understand the human microbiome and its association with disease [14, 15]. A lot of emphasis has been put on defining a healthy microbiome signature and core microbiome culminating in the Human Microbiome Project for cataloguing the microbial communities in different body sites. These projects have revealed that the gut microbiome is one of the most diverse and complex [14, 16‐18]. Although a core microbiome may exist every individual has a unique microbiota which is shaped by various parameters like genetic make-up, ethnicity, altitude, geographical location, mode of delivery and diet among others and also changes with age, travel, exposure to antibiotics and infections [14, 15, 19‐22] and onset of diseases [23].

The microbiota comprises archae, bacteria, viruses and unicellular eukaryotes. These carry out essential functions which are indispensible for maintaining a healthy state of the body and includes homeostasis, metabolism, immunity. This symbiotic association between the host and the microbiota is highly vulnerable as the fragile structure of the microbiota is prone to dysbiosis in the event of diseases. In the disease state the commensal flora is subdued by pathobionts (opportunistic pathogens and asymptomatically carried pathogens) [24]. The most common observation is Bacteroidetes/Firmicutes ratio which is high in the healthy state is reversed in the disease state with few exceptions [25]. Dysbiosis has been frequently studied in metabolic disorders [26], cancer [27], inflammatory diseases [28]. Specific microbes and specific signature of gut microbiota termed as enterotypes have been found to be associated with each of the diseases [29]. Only few studies have addressed gut microbiota dysbiosis in diarrhea. Most of these studies have been directed towards understanding dysbiosis in the event of infection by individual pathogens [30‐33] or in hospital acquired infections (HAIs) [34] or in Traveler’s diarrhea (TD) [25].

The current study is an unbiased pilot study conducted for characterizing the gut microbiota and the resistome from diarrheal stool and to see if we could find a statistically significant association of microbiota structure with diarrhea. We present the first comparative analysis of gut microbiota from twenty fecal samples collected from patients with symptoms of diarrhea. The stool samples were collected at the Infectious Diseases Beliaghata General Hospital (IDH) and Dr. B.C. Roy Memorial Hospital for Children (BCH), both in Kolkata, in Eastern India. These were subject to diagnostic test by classical microbiological method and were found to be associated with either distinct diarrheal etiology or with mixed infections and for some the etiology could not be determined by culture method currently deployed in our laboratory. Eastern India is endemic for diarrhea. Kolkata is a cosmopolitan city with a population of 5.8 million. It is the capital of the state of West Bengal (Fig. 1) and a major commercial hub of India where people of high, middle and low-income groups throng for job and business opportunities from across the country contributing to the remarkable cultural and ethnic diversity of the city. The Infectious Diseases and B.C. Roy Memorial Hospital in Kolkata has specialized facility for the treatment of diarrheal patients. It is the apex referral centre and sentinel surveillance centre for infectious diseases in West Bengal and Eastern India. Regular diarrheal stool collection takes place from the outpatient ward and from hospitalized patients. Therefore, NGS applied to study diversity of bacterial composition of the gut microbiome is anticipated to reveal striking biodiversity. The results could be a valuable resource for understanding the gut microbiota composition and resistome in the region. In our study we present the profile of the gut microbiota using 16S rRNA amplicon sequencing and resistome using whole genome shotgun (WGS) sequencing in diarrheal patients who were not subjected to any selective bias. They were randomly selected to represent the heterogeneity in a real community to catalogue the diversity of bacterial species present in the gut microbiota of the local community and in spite of observed inter-individual differences in enterotypes to define a shared microbiome. The study helped to understand the importance of the composition of the diarrheal gut microbiota that may be contributing to diarrheal pathogenesis, AMR and identify organisms that may be exploited to counterfeit the effect of diarrhea. The study helped to establish a catalogue of taxonomic units present in the gut microbiome of diarrheal subjects and to understand the superiority of WGS over 16S amplicon sequencing in studying the structure of the microbiota.

16S rRNA amplicon sequencing was used to study the gut microbiota associated with diarrhea in twenty diarrheal samples collected from two hospitals in Kolkata to define the core and variable microbiota in this part of India. Bacterial taxonomic identification was performed by matching DNA sequence homology of the metagenomic reads generated from 20 diarrheal samples to 1,45906 reference genomes available on the Genome Taxonomy Database [50]. We have been able to identify taxonomic units at different taxonomic levels namely, phyla, class, order, family, genera and species which were found in all the twenty diarrheal samples. Therefore, it may be inferred that these constituents may be present as part of the core microbiome in diarrheal patients. However, we cannot assert to what extent the proportion of these constituents has been altered or undergone dysbiosis compared to the normal or non-diarrheal microbiota, since, comparison of diarrheal and non-diarrheal stool samples was beyond the scope of our work. Next-generation sequencing is not easily accessible due to the constraints of expenses incurred in the sequencing and analysis process [55]. Therefore, we conducted a pilot study with a small sample size of 20 diarrheal fecal samples to determine microbiota composition during diarrhea and to define a bacterial signature for diarrhea, irrespective of the pathogen causing diarrhea.

We aimed to see differences in microbiota structure based on the diarrheal pathogen that was isolated by classical microbiological method. We found in diarrheal samples the dominant phylum is Firmicutes. In 11 out of 20 samples phylum Firmicutes was the most abundant phylum (Fig. 4). The Bacteroidetes/Firmicutes is an important indicator of bacterial dysbiosis [25]. The healthy gut has been found to have higher proportion of Bacteroidetes than Firmicutes [25]. 18 out of 20 diarrheal fecal samples showed higher abundance of Firmicutes than Bacteroidetes. From our study we conclude that the diarrheal gut has a higher abundance of Firmicutes than Bacteroidetes. Two samples S13 associated with V.cholerae O139 infection and S15 associated with Aeromonas sp. were found to have a higher proportion of Bacteroidetes compared to Firmicutes. In sample S13, which was obtained from an adult male of 40 years the dominant phylum was Bacteroidetes and in S15 obtained from a male child of 1 year old, Actinobacteria was the most abundant phylum. However, in this sample also Firmicutes was in higher abundance than Bacteroidetes. The gut microbiota primarily comprise Firmicutes, Bacteroidetes, Actinobacteria, Proteobacteria, Tenericutes and Fusobacteria [56]. The adult Microbiota is dominated by phyla Firmicutes, Bacteroidetes and Actinobacteria and depends on a host of intrinsic and extrinsic factors while the infant gut is usually dominated by phylum Actinobacteria [20, 56‐58]. In sample S16, which came from a female infant of 8 months old and Aeromonas sp. was the etiologic agent as confirmed by culture method, the most dominant phylum was found to be Firmicutes. Proteobacteria was the most abundant phylum in six samples. From all these samples V.cholerae was isolated by classical culture method (Table 1, Fig. 4). B/F ratio of diarrheal patients associated with V.cholerae infection was found to be statistically significant on the basis of one sample t and Wilcoxon test. This provided us with an insight into the B/F ratio that might be associated with cholera. The study provided us with a trend in microbiota structural composition in the diarrheal gut that could also be indicative of dysbiosis. However, comparison of the profile with that of non-diarrheal subjects will help in establishing the baseline of Bacteroidetes/Firmicutes ratio. This could be assertively used as an indicator for diarrhea.

Predominance of Proteobacteria and Firmicutes are indicators of a disturbed gut microflora [55]. 42 other phyla including Tenericutes, Fusobacteria and Candidate phylum radiation (CPR) were found in some of the samples. Their proportion was found to be very minute. Although Tenericutes and Fusobacteria have been shown to be a part of the core Microbiota [56], in our study these were absent in samples like S16, which was associated with Aeromonas sp. infection. Under the superphylum PVC [59] all phyla except Omnitrophica occurred in one or the other sample. These were Planctomycetes, Verrucomicrobiae, Chlamydiae and Lentisphaerae. Verrucomicrobiae associated with primarily beneficial bacteria and of environmental origin [59] was found to be in low abundance in most of the samples in which it occurred. Core Microbiota varies with geographic location, nationality and diet among other factors [20, 60]. Previous reports by other researchers in other parts of the world have shown the presence of Actinobacteria and Verrucomicrobia as dominant phyla in healthy subjects [20, 60]. A suppression in the proportion of these phyla in the diarrheal subjects in our study indicate either a characteristic of the Indian gut microbiota or dysbiosis associated with diarrhea. A study by Das et al. showed the healthy Indian gut consistently harbours 62% Firmicutes, 24% Bacteroidetes, 5.2% Actinobacteria and 4.2% Proteobacteria and a low abundance of Verrucomicrobia, Tenericutes and Fusobacteria were found in most of the individuals participating in the study [61]. In the present study we found 38% Firmicutes, 10% Bacteroidetes, 12% Actinobacteria and 19% Proteobacteria. The difference in the abundance of these phyla as observed in the present study could be due to diarrhea and diet, ethnicity, geographical location and other environmental factors influencing the proportion of these constituents. A study conducted by Monira et al. addressing the gut microbiota composition in healthy and malnourished children in Bangladesh showed that in healthy children Proteobacteria and Bacteroidetes accounted for 5% and 44% respectively [62]. As Eastern India and Bangladesh are comparable demographies we may assume that the lower abundance of Proteobacteria in the healthy gut observed in Bangladeshi children has been altered in diarrheal subjects resulting in higher abundance of Proteobacteria in the diarrheal subjects of the present study.

Candidate Phyla Radiation (CPR) like Candidatus Falkowbacteria, Candidatus Moranbacteria and others, belonging to the Parcubacteria group (Table 2) were found in our study. These are uncultured bacteria of environmental origin and involved in important ecological activity like sulfur-reduction and other biogeochemical cycles like carbon and hydrogen cycles [63, 64]. These are of ancient lineage, mostly symbionts or episymbionts, lack biosynthetic pathways and have not been cultured due to their stringent metabolism [64]. Our metagenomic data showed the presence of Candidatus Saccharibacteria or TM7 phylum which has been found to be a potential pathogen with a parasitic lifestyle, associated with human inflammatory mucosal diseases and often recovered from wastewater and clinical environments [65, 66].

The uncultivated Candidate phyla is referred to as “microbial dark matter” [66]. Its presence in diarrheal samples from patients in and around Kolkata is a matter of concern about environmental pollution and intestinal colonization of organisms with pathogenic potential. It will be interesting to investigate how they have adapted to the intestinal habitat and about the transmission of these organisms into the host from the environment. In the future it will be interesting to look for these metagenomes in healthy/non diarrheal microbiome.

In the recent years a large number of published reports attempting to define the core gut microbiome of Indians are available [61, 67]. Bacterial composition at the genus level has been found to be influenced by location and diet [61]. Kulkarni et al. showed the presence of Prevotella sp., Bacteroides sp., Megasphaera sp., Roseburia sp., from fecal samples of 43 Indians. Das et al., showed the presence of a core microbiota comprising 54 genera from fecal samples of individuals from rural, urban and high-altitude dwellers in India [61]. Another study conducted by Lin et al. showed that healthy Bangladeshi chidren harboured more of Prevotella, Butyrivibrio, and Oscillospira and were depleted in Bacteroides [68]. Our study is the first attempt to present a core microbiota signature in diarrheal subjects from Eastern India. We found 18 genera that were present in all the 20 samples (Table 3). Prevotella sp. was absent in S10. The diarrheal etiology of this sample could not be successfully determined by culture method. This sample was from a 55 year old male from Kolkata who was hospitalized for 1 day at ID Hospital in Kolkata. Prevotella sp. has been found to be associated with the core human gut microbiome [61]. It is a pathobiont of clinical significance. A positive correlation between the upsurge of Prevotella copri and diarrhea has been estimated by previous studies [69].

The study showed the presence of commensals, pathobionts and pathogenic bacteria in the diarrheal gut microbiome. Pathobionts may cause inflammatory disorders or may cause infections in the event of compromised immunity [70, 71]. The presence of pathogens like V. cholerae, Helicobacter pylori etc. in addition to the etiologic agent isolated by culture was found in many samples. This is a matter of grave concern as asymptomatic carriers act as reservoirs of infections and expedite the transmission of infections. Commensals like Bifidobacterium sp., Ruminococcus sp., Fecalibacterium sp., Lactobacillus sp., Lactococcus sp., were found in the present study and are intrinsic colonizers of the human gut [72]. Commensals play a protective role by mediating colonization resistance and preventing colonization by pathogens and opportunistic pathogens, prevent intestinal barrier impairment and suppresses pro-inflammatory factors thereby preventing diarrhea [73, 74]. Earlier studies showed that abundance of certain commensals remained unchanged before, during and after recovery from acute diarrhea in children while others like Eubacterium sp., Fecalibacterium sp., Prevotella sp., Bacteroides sp., showed marked differences during acute diarrhea and after recovery [75]. It will be interesting to investigate whether the reduction in proportion of commensals prior to diarrheal onset lay the ground for diarrheal pathogenesis. Previous studies on diarrhea associated microbiota had found a positive correlation between diarrhea and pathogenic bacteria like Eschericia sp., Shigella sp., Granulicatella sp, Streptococcus sp. [76], We found the existence of these pathogenic genera in our study subjects. Eschericia sp. was not found in S5, S7, S9, S12, S14, S18 and all of these samples were associated with V. cholerae infection. These findings led us to test if there is a correlation in the relative abundance of various families of pathogens and among pathogens and commensals which could be of significance for diarrheal etiology or bear implications for diarrheal treatment.

We found an association among the relative abundance of families of commensals and pathogens (Fig. 6a). Although some of the associations were not statistically significant it succeeded to present a trend which may be useful for understanding the agonistic and antagonistic relationship among these families and could show direction in preventive and therapeutic modules of diarrheal diseases. These correlation could become statistically significant if performed on a larger sample size. We found that the commensals Bifidobacteriaceae and Lachnospiraceae were negatively correlated with pathogens Enterobacteriaceae and Vibrionaceae. Among the pathogenic groups, family Enterobacteriaceae was higher than both Vibrionaceae and Aeromonadaceae thereby shedding light on the trend observed in gut microbiota during diarrhea. Streptococcaceae and Enterobacteriaceae were positively correlated indicating that these two pathogenic groups show the same trend in gut microbiota structural composition in diarrhea. Enterobacteriaceae are a family of potential pathogens and our study showed that these outnumber other families of potential pathogens like Vibrionaceae and Aeromonadaceae in diarrhea implying the obvious trend in diarrheal dysbiosis.

We observed differences in relative abundance among various families of bacteria in samples found to be associated with V. cholerae or Aeromonas sp. as etiologic agents. These are two common diarrheal pathogens and we wanted to examine if we could derive any significant association of any pathogenic or commensal family with these specific diarrheal etiology. We noted a trend in the difference in abundance of Vibrionaceae with Bifidobacteriaceae and Lachnospiraceae and Enterobacteriaceae. Vibrionaceae was higher than the commensals while lower than Enterobacteriaceae. Aeromonadaceae abundance was lower than those of the commensals and Enterobacteriaceae but the difference was non-significant. These findings suggest that in diarrhea commensals are suppressed by pathogens belonging to these families and could bear implications for probiotic therapy in diarrhea with commensal gut pathogens. This is also suggestive of the pattern of dysbiosis occurring in diarrhea. The same comparative analysis if performed in a healthy study cohort may help to determine if the observed differences in our analysis is due to dysbiosis associated with diarrhea.

Mean abundance of Enterobacteriaceae was significantly higher than that of Aeromonadaceae in the diarrheal study cohort. Significant difference in mean abundance among Bacteroidaceae, Enterobacteriaceae and Vibrionaceae was observed. These findings suggest that in diarrhea certain families of pathogens overpower others and this may lead to co-infections, co-morbidities leading to complications in diarrheal treatment.

Statistically significant positive correlation was observed among the families of commensals like Bifidobacteriaceae, Lachnospiraceae and Ruminococcaceae, indicating agonistic relationship among these and significant negative correlation among families of commensals and pathogens like Enterobacteriaceae and Lachnospiraceae and Enterobacteriaceae and Ruminococcaceae were observed. All these observations indicate antagonistic relationship bearing promise of future exploitation of these tendencies for development of probiotics.

The samples had a variable range of α-diversity. S1 had the least while S20 had the maximum diversity. Samples like S1 and S20 associated with the same diarrheal etiologic agent, VC, had stark differences in Shannon-indices indicating that other parameters are crucial for microbiota structural composition. For analysis of β-diversity samples were grouped according to diarrheal agent isolated from it by culture method. The samples did not group into clusters based on the etiologic agent. We anticipate this was due to the small sample size and also factors other than the etiologic agent of diarrhea determining the bacterial composition in the gut.

The gut of diarrheal patients carries a high abundance of antimicrobial resistance genes (ARGs) and the members of the microbiota have been found to carry these genes in their genomes and act as reservoirs of AMR in the gut [55, 77]. We used WGS to sequence five diarrheal samples to study the resistome and understand the origin of ARGs in the gut microbiome. We selected the fecal samples to see if variation in these aspects existed based on demography, etiology and α-diversity. In spite of the differences in demography, etiology and α-diversity all the samples showed the presence of the four classes of ARGS namely, tetracyclines, β-lactams, aminoglycosides and macrolides. Even though samples like S1 and S2 were associated with the same diarrheal etiology V. cholerae and were from the same district, 24 Parganas, their resistome analysis revealed difference in relative abundance of the same ARGs like tetracyclines, quinolones, β-lactams, aminoglycosides and macrolides. Although S1 had the lowest α-diversity, it did not have the lowest diversity of ARGs although had the lowest number of total ARGs compared to the others. S9 and S10 were both from Kolkata but S10 had the highest number of ARGs while S9 had much lower number of ARGs and S10 had much higher relative abundance of each class of ARGs compared to S9. Moreover, quinolones were absent in S9. We conclude that in this region Eschericia sp. is the major contributor of ARGs in the gut. This is of grave concern. Eschericia sp. includes both commensals and pathogens. They are involved in metabolism and defense mechanisms [78]. Eschericia sp. are resident microbes of the gut. These will act as reservoirs for dissemination of ARGs into other bacteria in close proximity. Moreover, from the five fecal samples genomes of E. coli D, E. marmotae, E. albertii, E. fergusonii were reconstructed in addition to others (Fig. 13). Many of these pathogens are MDR as confirmed by previous studies [79].

We found a high abundance of resistance against tetracyclines, macrolides, aminoglycosides, quinolones and β-lactams. This presents a menacing picture of the AMR crisis in countries like India. These are last resort drugs against enteric pathogens like E.coli, K. pneumoniae, V. cholerae which are common diarrheal pathogens in India. Our study revealed that resistance determinants against the most important classes of antimicrobials are present in the gut of people residing in this region. This will contribute to transmission and spread to the community and the environment and lead to the emergence of MDR and XDR (Extensively Drug Resistant) strains.

Diarrhea is associated with dysbiosis of microbiota [75]. The dynamics of gut microbiota has been well-studied in case of invading pathogens like V.cholerae [80]. We used NGS to study the gut microbiota in acute diarrheal patients in the present study. The results showed that a core microbiota exists in diarrheal patients. Specific signature of microbiota composition corresponding to distinct diarrheal etiology could not be established. We anticipate it is due to the small sample size. The trend that we observed can be confirmed by expanding the sample size in the future. The study helped to reveal the critically high abundance of AMR determinants against the most crucial drugs administered for diarrheal treatment and confirmed the existence of these determinants in the gut of diarrheal patients and originating from genomes of pathogens residing in the gut. From these ecological cross-talk future threat of infections by MDR bacteria would emanate. The study highlights the presence of asymptomatic carriers of pathogens who are serving as reservoirs of important infectious agents and expediting community transmission of diarrheal pathogens.

In the study two NGS techniques were used simultaneously. 16S rRNA amplicon sequencing helps to identify bacterial taxa but not function. WGS provides comprehensive information about both the structure and function of the microbiota. It also helps to identify the genomes contributing to those functions. Therefore we conclude that if molecular epidemiological laboratories can overcome financial constraints, WGS would be the preferred technique for investigating the constituent genomes of the microbiome and annotate their functional role.

The pilot study revealed significant antagonistic correlation of families of commensals like Lachnospiraceae and Ruminococcaceae with pathogens like Enterobacteriaceae, on the basis of Spearman’s correlation coefficient test. Bacteria with probiotic capability can be identified and these can be developed as probitocs for alternative therapy to replace or supplement antibiotic therapy in diarrhea. The existence of “microbial dark matter” in diarrheal gut evident from our study is indicative of contamination of the gut microbiota with rare and dangerous bacteria. This would help in epidemiological analysis to trace the origin and understand the route of transmission of members of Candidate phyla into the diarrheal gut microbiome. Consequently, it will be useful to reduce the occurrence of such organisms in the environment and the gut. Overall, the study on metagenomic sequencing of diarrheal microbiome is the first of its kind, from Eastern India revealing the core and variable microbiota associated with diarrhea and has immense implications for understanding diarrheal etiology.

16S V3–V4 metagenome libraries were prepared using region-specific primers. DNA samples were loaded on gel to examine the bands followed by 0.7 × Hiprep bead clean up using HighPrep™ clean up system (catalogue numberAC-60050, MagBio genomics Inc., USA) to avoid impurities and amplified for 26 cycles of round 1 PCR using KAPA HiFi Hot-Start PCR Kit (catalogue number KM2602, KAPA Biosystems Inc., Boston, MA, USA). The forward and reverse primer concentration was kept at 5 µM each. The amplicons were analyzed on 1.2% agarose gel. 1 µl of diluted round 1 PCR amplicons were used for Indexing PCR (Round 2). Round 1 PCR amplicons were amplified for 10 cycles to add Illumina sequencing barcoded adaptors (Nextera XT v2 Index Kit, catalogue number FC-131-1002 Illumina Inc., CA, USA). Illumina Adapter Sequences used were: 5′-AATGATACGGCGACCACCGAGATCTACAC[i5]TCGTCGGCAGCGTC and 5′-CAAGCAGAAGACGGCATACGAGAT[i7] GTCTCGTGGGCTCGG where [i5, i7] are unique dual index sequences to identify sample-specific sequencing data.

Round 2 PCR amplicons (sequencing libraries) were analyzed on 1.2 percent agarose gel, cleaned using HighPrep™ clean up system and quality checked. The library was diluted to 4 nM using 10 mM Tris (pH 8.5) and 5 µl of each library was aliquotted and mixed to pool the libraries. The pooled library was denatured by addition of NaOH followed by heat denaturation and the DNA samples were diluted and finally loaded onto the Illumina MiSeq system and sequencing was performed to generate (300*2) V3–V4 paired-end reads.

The Illumina paired end V3–V4 raw reads (300*2) were submitted to the European Nucleotide Archive (ENA) for validation and further analysis using the MGnify pipeline provided by the EMBL server. The study was assigned the number MGYS00005131. The raw reads were processed using MGnify v4.1. SeqPrep [36] was used to merge the overlapping raw reads into a single longer read. Trimmomatic [37] and Biopython [38] were used to trim and filter these initial reads by removing > 10% undetermined nucleotides and adapter sequences and filtering out < 100 bp long sequences to generate processed reads which were annotated using MAPseq [39] framework for taxonomic classification and Operational Taxonomic Unit (OTU) mapping. For classification of OTUs, paired-end reads with > 97% sequence similarity were considered. The sequences of raw and processed reads can be accessed through the EMBL server with the accession number MGYS00005131.For multivariate analysis and graphical representation of the metadata tools Codaseq [40], Vegan [41] and Ape [42] on the Phyloseq [42] package, ggplot2 [43] on R Studio (R studio Inc, Boston, MA, USA) were used. Biom files generated by MAPseq in the MGnify pipeline were imported into R package. Principal component Analysis (PCA) was performed using the Phyloseq package for analysis of abundance of OTUs and entitities within different taxonomic ranks namely, phylum, class, order, family, genera and species. Abundance was expressed as percentage. Relative abundance of different entities within a taxonomic level was represented as histogram to show taxonomic diversity and abundance. 0–5 percent was used as the threshold. The top fifteen to twenty-five OTUs within each taxon were plotted for each sample. α-diversity was calculated to estimate species richness and evenness of each sample. Accordingly, OTUs were rarefied at even depth and Shannon index was calculated. To calculate β-diversity between the samples ordination was performed and principal coordinates plots were generated based on pairwise weighted UniFrac distances. Pie, bar, stacked and interactive krona charts were generated by the taxonomic analysis steps of the MGnify v 4.1 pipeline. Bacteroidetes/Firmicutes ratio was calculated and compared among the 20 samples. For bivariate analysis normal distribution, z-score, unpaired t-tests, ANOVA (Analysis of Variance) and one sample t and Wilcoxon tests were calculated to represent the statistical significance of the taxonomic composition and abundance data. Correlation coefficient using Spearman’s rank co-efficient test was used to study correlation among abundance of families Bifidobacteriaceae, Lachnospiraceae, Ruminococcaceae, Enterobacteriaceae, Vibrionaceae, Streptococcaceae to derive if any significant association existed among them in diarrhea. Kruskal–wallis test was used to compare abundance of three families of commensal bacteria namely Bifidobacteriaceea, Lachnospiraceae and Ruminococcaceae and three families of pathogenic bacteria namely Enterobacteriaceae, Bacteroidaceae and Vibrionaceae to see if an association could be established among the relative abundance of these families which could have a significance for diarrheal etiology.

Unpaired t-test was used to compare difference in relative abundance of Bifidobacteriaceae with Enterobacteriaceae and Vibrionaceae, Lachnospiraceae with Enterobacteriaceae and Vibrionaceae, also between Enterobacteriaceae and Vibrionaceae and between Enterobacteriaceae and Aeromonadaceae in diarrhea and calculations were based on 20 samples. Wilcoxon matched-pairs signed rank test was used to compare difference in relative abundance of Vibrionaceae with Bifidobacteriaceae, Enterobacteriaceae, Lachnospiraceae and Ruminococcaceae in samples diagnosed with Vibrio sp. by culture method. Unpaired t-test was used to compare differences in relative abundance of Aeromonadaceae with Bifidobacteriaceae, Enterobacteriaceae, Lachnospiraceae in samples diagnosed with Aeromonas sp. by culture method.

Figure 2 presents the workflow of library preparation, sequencing and metagenomic analysis.